Bandwidth-efficient Inference for Neural Image Compression
Abstract
With neural networks growing deeper and feature maps growing larger, limited communication bandwidth with external memory (or DRAM) and power constraints become a bottleneck in implementing network inference on mobile and edge devices. In this paper, we propose an end-to-end differentiable bandwidth efficient neural inference method with the activation compressed by neural data compression method. Specifically, we propose a transform-quantization-entropy coding pipeline for activation compression with symmetric exponential Golomb coding and a data-dependent Gaussian entropy model for arithmetic coding. Optimized with existing model quantization methods, low-level task of image compression can achieve up to 19x bandwidth reduction with 6.21x energy saving.
Cite
@article{arxiv.2309.02855,
title = {Bandwidth-efficient Inference for Neural Image Compression},
author = {Shanzhi Yin and Tongda Xu and Yongsheng Liang and Yuanyuan Wang and Yanghao Li and Yan Wang and Jingjing Liu},
journal= {arXiv preprint arXiv:2309.02855},
year = {2023}
}
Comments
9 pages, 6 figures, submitted to ICASSP 2024